Open Cloud Assistant connects AI agents to real services, files, personal memory, messaging platforms, and coding tools.

Treat deployment credentials and personal runtime state as production secrets.

• API keys or provider tokens
• OAuth credentials
• bearer tokens or JWTs
• SSH private keys
• .env files
• messaging credentials
• personal memory or private context
• conversation history
• runtime databases
• authentication state
• session identifiers
• production logs containing private prompts or secrets

Automated coding agents should operate with restricted permissions.

They should not receive unrestricted access to:

• Git push credentials
• secret files
• personal-memory stores
• unrelated filesystem locations
• production databases
• system service control
• arbitrary network commands

A trusted outer workflow should create backups, validate changes, and perform privileged repository operations.

The public repository must be built from reusable source, templates, documentation, and sanitized integration logic.

Personal deployments must keep private runtime state outside Git.

Before a public release:

1. run the public credential/privacy audit,
2. run smoke and integration tests,
3. verify third-party licenses and notices,
4. test a clean supported installation.

Do not open a public issue containing a real credential, private memory, or authentication artifact.

If a credential was exposed, revoke or rotate it immediately before reporting the issue.